[PATCH] [PR c++/84610,84642] recover from implicit template parms gracefully

[Jason Merrill]

On Tue, Mar 20, 2018 at 5:57 PM, Alexandre Oliva <aoliva@redhat.com> wrote:
> On Mar 20, 2018, Jason Merrill <jason@redhat.com> wrote:
>
>>> While debugging this, I first tried another patch, that avoids the same
>>> ICEs.  I thought this one was a more complete solution, and it renders
>>> the other unnecessary, but I still though it might be useful to disable
>>> auto->implicit_parm while parsing declarators, so as to avoid useless
>>> processing.
>
>> Better, I think, to disable it while parsing attributes; we could
>> still encounter auto as a template argument in a declarator.
>
> How about this?  Regstrapping, ok to install if it passes?
>
> Disable auto_is_implicit_function_template_parm_p while parsing attributes
>
> for  gcc/cp/ChangeLog
>
>         PR c++/84610
>         PR c++/84642
>         * parser.c (temp_override): New template class,
>         generalizing...
>         (cp_parser_parameter_declaration_clause): ... this cleanup.
>         Use it.
>         (cp_parser_gnu_attributes_opt): Use it.
>         (cp_parser_std_attribute): Use it.
> ---
>  gcc/cp/parser.c |   36 ++++++++++++++++++++++++++----------
>  1 file changed, 26 insertions(+), 10 deletions(-)
>
> diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
> index 5cc201604166..ce05615adfba 100644
> --- a/gcc/cp/parser.c
> +++ b/gcc/cp/parser.c
> @@ -255,6 +255,21 @@ static bool cp_parser_omp_declare_reduction_exprs
>  static void cp_finalize_oacc_routine
>    (cp_parser *, tree, bool);
>
> +template <typename T>
> +class temp_override
> +{
> +  T& overridden_variable;
> +  T saved_value;
> +public:
> +  temp_override(T& var) : overridden_variable (var), saved_value (var) {}
> +  temp_override(T& var, T overrider)
> +    : overridden_variable (var), saved_value (var)
> +  {
> +    overridden_variable = overrider;
> +  }
> +  ~temp_override() { overridden_variable = saved_value; }
> +};

Let's put this in cp-tree.h, with warning_sentinel.

> +  (void)cleanup;

There are lots of RAII variables without this explicit cast to void,
and they don't seem to have been a problem; let's drop it here as
well.

Jason